Lobsters extract information from complex signals in turbulent odor plumes and it guides them to mates or food sources. To test hypotheses about this guidance information, we have developed a robot as a physical model of a lobster. Here we present the results of experiments designed to test the efficacy of amplitude information-a single component of a complex signal-in guidance. The robot used a bilateral pair of conductivity sensors (sensor surface spacing = 5-7 cm, 5 cm separating two l-cm wide sensors) to sense a salt plume simulating an odor plume. The experiments were performed in a fresh-water flume with a mean flow rate of0.6 cm/s. A 0.76 MNaCl solution (containing crystal violet for visualization and ethanol to adjust buoyancy) was injected parallel to the flow from a 2 mm diameter nozzle into the flume at a rate of 250 ml/min. The resulting plume had two distinct regions: a proximal cone originating at the source, and a distal patch field downstream from the jet. The proximal jet is the region where the velocity of the jet exceeds the mean flow in the flume. The distal patch field corresponds to plume positions downstream from the proximal cone where the mean flow is the major source of plume velocity (relative to the floor). Two robot control algorithms were tested: 1. The robot turns toward the side with the higher salt conductance signal or goes forward if the difference between the right and left sensor signals drop below 9 FS. 2. As in # 1, with the added feature that the robot goes backward if the conductances of both sensors drop below a threshold of 7 ps. The robot was placed in the center of the flume, 90 cm downstream from the plume source, and was started in two orientations for each algorithm: pointed upstream directly into the oncoming plume, and pointed 45 degrees to the right of the plume axis. Each of the four conditions (2 orientations and 2 algorithms) was replicated 10 times. The robot’s trajectory was recorded by a video camera. Data from a single run using algorithm #l are presented in Figure 1. As the robot moved through the patch field, its behavior was characterized by sequences of abrupt, brief turns that occurred at irregular intervals. When it entered the proximal jet, the robot moved with more regular side-to-side oscillations: a characteristic series of alternating smooth left and right turns (often of greater magnitude than those seen in the distal patch field). Once inside the proximal jet the robot often found its way to the source (50% algorithm #1 and 72% algorithm #2). The starting orientation had a substantial effect on the success of the algorithms. Algorithm #2 with the robot pointing into the plume had a higher rate of direct “hits” onto the source than algorithm #l with the same orientation (66% vs 33%). We attribute the greater failure rate of algorithm #I to the fact that when both sensors happen to exit the plume, algorithm #l moves the robot in a straight line away from the point of exit. The Right Sensor A Llll ae:IlJ”l (inverted)